Stage Positioning Repeatability Testing to ISO 230-2: Running the Protocol So the Numbers Hold Up

The test structure: targets, approaches, statistics

The ISO 230-2 cycle is simple to state and demanding to execute honestly. The axis visits each of at least five target positions along its travel, approaching each target a stated number of times from each direction—five is the convention—so the sequence alternates direction as it works through targets. At every arrival, the actual position is recorded by the external measuring instrument. The raw data set is therefore a matrix: targets by approaches by directions.

From each target's directional subset the statistics fall out in a fixed order. For each direction, the mean position and the standard deviation of the approaches give the unidirectional repeatability at that target. The difference between the two directional means at the same target is the systematic reversal offset. Combining both directions yields the bidirectional repeatability at that target, which the standard then expresses as an axis-level value. The mean deviations across all targets give the positioning accuracy of the axis. Every one of these numbers is defined over the whole set of targets; the worst target and the average target are both information, and a report that gives only one of them is making a choice it should disclose.

The reason the standard insists on both directions kept separate is that they answer different engineering questions. A unidirectional repeatability number describes a machine whose process always approaches the work from the same side—useful and real. A bidirectional number describes everything else, and it is always larger, because it includes the reversal offset. Comparing a unidirectional datasheet figure against a competitor's bidirectional figure is the single most common way stage specifications mislead buyers.

Setup conditions that decide whether the result means anything

The instrument setup comes first. Laser interferometer optics are aligned to the axis of motion with Abbe offset minimized, the wavelength is compensated for air temperature, pressure and humidity through the environmental unit, and the optics warm up with the laser before data collection. The interferometer measures displacement between its reference and moving optics—any rotation, Abbé arm or deadpath error in the setup lands directly in the numbers, so alignment quality is checked before the first cycle rather than discovered in the results.

The thermal state is the second decisive condition. The machine runs its specified warm-up before the test; ambient temperature is recorded and held reasonably stable during the run; the thermal state at test time is written into the record. An uncompensated stage drifting a few microns over a cold-to-warm transition will produce scatter that has nothing to do with its mechanical repeatability, and a stage tested cold then used warm will appear to have 'lost' repeatability it never actually had in that state.

The motion program is the third. Approach speed, dwell before the position reading, and the final creep into target must match what the controller does in production, because repeatability is a property of the machine as commanded, not of the axis in the abstract. A test that creeps in slowly at a fraction of production speed flatters the axis; a dwell too short captures a still-settling position and inflates scatter. Both the values and the conditions that produced them belong in the report.

• Optics aligned with minimal Abbe offset; deadpath and cosine error corrected
• Environmental compensation active; air temperature, pressure and humidity logged
• Warm-up completed; thermal state stated in the record
• Approach speed, dwell and creep matched to production motion

Reading the results as an engineering document

A complete ISO 230-2 report is a table, not a sentence. Per-target unidirectional and bidirectional repeatability, the reversal offsets, the mean deviations and the axis-level summary values appear together with the raw scatter they came from. The per-target columns carry the diagnostic value: scatter that grows toward one end of travel suggests mechanical or scale issues specific to that region; a reversal offset that varies across travel points to different systematic contributors than a constant one; a single outlier target usually indicates a local event—a bump in the scale, a guide defect—rather than a globally noisy axis.

Repeatability then has to be placed in the machine's error budget. It is one term among many: Abbe effects, straightness and angular errors, thermal behavior, and the interaction with the process all contribute to the accuracy the workpiece actually sees. A stage with excellent repeatability and an uncompensated Abbe arm still produces inaccurate parts. The budget allocates which term dominates which tolerance, and the ISO 230-2 results are the measured anchor for the repeatability share of that allocation.

Writing repeatability into acceptance and purchase documents

A purchasing specification that simply says 'repeatability ±X µm' invites the ambiguity this guide exists to close. The document should state the standard and which statistic is meant—unidirectional or bidirectional, per-target worst value or axis-level value—that both approach directions are tested, the number of targets and cycles, the warm-up and thermal conditions, and the instrument with its uncertainty. With those terms fixed, two suppliers' numbers become comparable and the acceptance test becomes a replay of a defined procedure instead of a negotiation.

The same terms carry into routine verification. Periodic abbreviated runs—the same targets, the same directions, a reduced cycle count—trend the axis's repeatability against its acceptance baseline and distinguish mechanical degradation from thermal or process drift. Repeatability tested once at handover and never again is a snapshot; repeatability tested to a fixed protocol over life is a health record, and it is the second kind that keeps production yields explainable.